Fill the right column of a matplotlib legend first

Question

Hey I am trying to fit a legend onto a plot so that it doesn't obscure the graph.

import numpy as np
import matplotlib.pyplot as plt

X = np.linspace(0,100,11)

plt.plot(X,-X, label='plot 1')
plt.plot(X,-2*X, label='plot 2')
plt.plot(X,-3*X, label='plot 3')

leg=plt.legend(ncol=2)
leg.get_frame().set_visible(False)

plt.show()

So in the minimum working example, above, what I want to be able to do is move the 'plot 2' label in the legend into the right column, i.e. directly under 'plot 3'.

Any help would be appreciated, thanks.

Answer 1

The legend is filling in the columns from left to right. In other words, if you trick it into believing that there is another line (without any text or line color in the legend), then you can populate the space under 'plot 3'.

import numpy as np
import matplotlib.pyplot as plt
from pylab import *

X = np.linspace(0,100,11)

plt.plot(X,-X, label='plot 1', color='red')
plt.plot(X,-2*X, label='plot 2', color='green')
plt.plot(X,-3*X, label='plot 3', color='blue')


line1 = Line2D(range(10), range(10), marker='', color="red")
line2 = Line2D(range(10), range(10), marker='',color="green")
line3 = Line2D(range(10), range(10), marker='', color="blue")
line4 = Line2D(range(10), range(10), marker='', color="white")
plt.legend((line1,line4, line3,line2),('plot1','','plot3','plot2'),numpoints=1, loc=4,ncol=2)

plt.show()

Answer 2

A diffrerent implementation of @cosmosis answer. It's probably more flexible.

import numpy as np
import matplotlib.pyplot as plt

X = np.linspace(0,100,11)

plt.plot(X,-X, label='plot 1', color='red')
plt.plot(X,-2*X, label='plot 2', color='green')
plt.plot(X,-3*X, label='plot 3', color='blue')

(lines, labels) = plt.gca().get_legend_handles_labels()
#it's safer to use linestyle='none' and marker='none' that setting the color to white
#should be invisible whatever is the background
lines.insert(1, plt.Line2D(X,X, linestyle='none', marker='none'))
labels.insert(1,'')

plt.legend(lines,labels,numpoints=1, loc=4,ncol=1)

plt.show()

An other option is to create two legends as here and then displace them using bbox_to_anchor keyword here

(lines, labels) = plt.gca().get_legend_handles_labels()
leg1 = plt.legend(lines[:1], labels[:1], bbox_to_anchor=(0,0,0.8,1), loc=1)
leg2 = plt.legend(lines[1:], labels[1:], bbox_to_anchor=(0,0,1,1), loc=1)
gca().add_artist(leg1)

doing it I get

without the need of add any on other object.

Answer 3

A third approach, based on the answer of Franesco. Draw a placeholder line with alpha=0 (transparent).

#draw your actual lines here
#plt....

lines, labels = plt.gca().get_legend_handles_labels()

lines.insert(1, plt.Line2D([],[], alpha=0))
labels.insert(1,'')

plt.legend(lines,labels,ncol=2)
plt.show()

This has the same advantages as described in his answer. However that solution did not work in matplotlib 1.5.1 (under python3) saying ValueError: Unrecognized marker style none